It's also what makes it possible for you to write, say, a high-performance floating point matrix multiplication function and have the compiler emit code accepts

It's also what makes it possible for you to write, say, a high-performance floating point matrix multiplication function and have the compiler emit code accepts floating point inputs without having to consider the possibility that the arguments were supplied as strings.